[PATCH v3 1/2] compat: add a mkstemps() compatibility function

[PATCH v3 1/2] compat: add a mkstemps() compatibility function

[David Aguilar]

mkstemps() is a BSD extension so provide an implementation
for cross-platform use.

Signed-off-by: David Aguilar <davvid@gmail.com>
Tested-by: Johannes Sixt <j6t@kdbg.org> (Windows)
---
 Makefile          |   19 +++++++++++++++
 compat/mkstemps.c |   67 +++++++++++++++++++++++++++++++++++++++++++++++++++++
 config.mak.in     |    1 +
 configure.ac      |    6 ++++
 git-compat-util.h |    5 ++++
 5 files changed, 98 insertions(+), 0 deletions(-)
 create mode 100644 compat/mkstemps.c

diff --git a/Makefile b/Makefile
index eaae45d..a70b5f0 100644
--- a/Makefile
+++ b/Makefile
@@ -52,6 +52,8 @@ all::
 #
 # Define NO_MKDTEMP if you don't have mkdtemp in the C library.
 #
+# Define NO_MKSTEMPS if you don't have mkstemps in the C library.
+#
 # Define NO_SYS_SELECT_H if you don't have sys/select.h.
 #
 # Define NO_SYMLINK_HEAD if you never want .git/HEAD to be a symbolic link.
@@ -636,10 +638,12 @@ EXTLIBS =
 
 ifeq ($(uname_S),Linux)
 	NO_STRLCPY = YesPlease
+	NO_MKSTEMPS = YesPlease
 	THREADED_DELTA_SEARCH = YesPlease
 endif
 ifeq ($(uname_S),GNU/kFreeBSD)
 	NO_STRLCPY = YesPlease
+	NO_MKSTEMPS = YesPlease
 	THREADED_DELTA_SEARCH = YesPlease
 endif
 ifeq ($(uname_S),UnixWare)
@@ -651,6 +655,7 @@ ifeq ($(uname_S),UnixWare)
 	SHELL_PATH = /usr/local/bin/bash
 	NO_IPV6 = YesPlease
 	NO_HSTRERROR = YesPlease
+	NO_MKSTEMPS = YesPlease
 	BASIC_CFLAGS += -Kthread
 	BASIC_CFLAGS += -I/usr/local/include
 	BASIC_LDFLAGS += -L/usr/local/lib
@@ -674,6 +679,7 @@ ifeq ($(uname_S),SCO_SV)
 	SHELL_PATH = /usr/bin/bash
 	NO_IPV6 = YesPlease
 	NO_HSTRERROR = YesPlease
+	NO_MKSTEMPS = YesPlease
 	BASIC_CFLAGS += -I/usr/local/include
 	BASIC_LDFLAGS += -L/usr/local/lib
 	NO_STRCASESTR = YesPlease
@@ -702,6 +708,7 @@ ifeq ($(uname_S),SunOS)
 	NO_MEMMEM = YesPlease
 	NO_HSTRERROR = YesPlease
 	NO_MKDTEMP = YesPlease
+	NO_MKSTEMPS = YesPlease
 	OLD_ICONV = UnfortunatelyYes
 	ifeq ($(uname_R),5.8)
 		NO_UNSETENV = YesPlease
@@ -724,6 +731,7 @@ ifeq ($(uname_O),Cygwin)
 	NO_D_INO_IN_DIRENT = YesPlease
 	NO_STRCASESTR = YesPlease
 	NO_MEMMEM = YesPlease
+	NO_MKSTEMPS = YesPlease
 	NO_SYMLINK_HEAD = YesPlease
 	NEEDS_LIBICONV = YesPlease
 	NO_FAST_WORKING_DIRECTORY = UnfortunatelyYes
@@ -767,11 +775,13 @@ ifeq ($(uname_S),NetBSD)
 	BASIC_LDFLAGS += -L/usr/pkg/lib $(CC_LD_DYNPATH)/usr/pkg/lib
 	THREADED_DELTA_SEARCH = YesPlease
 	USE_ST_TIMESPEC = YesPlease
+	NO_MKSTEMPS = YesPlease
 endif
 ifeq ($(uname_S),AIX)
 	NO_STRCASESTR=YesPlease
 	NO_MEMMEM = YesPlease
 	NO_MKDTEMP = YesPlease
+	NO_MKSTEMPS = YesPlease
 	NO_STRLCPY = YesPlease
 	NO_NSEC = YesPlease
 	FREAD_READS_DIRECTORIES = UnfortunatelyYes
@@ -787,12 +797,14 @@ endif
 ifeq ($(uname_S),GNU)
 	# GNU/Hurd
 	NO_STRLCPY=YesPlease
+	NO_MKSTEMPS = YesPlease
 endif
 ifeq ($(uname_S),IRIX64)
 	NO_IPV6=YesPlease
 	NO_SETENV=YesPlease
 	NO_STRCASESTR=YesPlease
 	NO_MEMMEM = YesPlease
+	NO_MKSTEMPS = YesPlease
 	NO_STRLCPY = YesPlease
 	NO_SOCKADDR_STORAGE=YesPlease
 	SHELL_PATH=/usr/gnu/bin/bash
@@ -805,6 +817,7 @@ ifeq ($(uname_S),HP-UX)
 	NO_SETENV=YesPlease
 	NO_STRCASESTR=YesPlease
 	NO_MEMMEM = YesPlease
+	NO_MKSTEMPS = YesPlease
 	NO_STRLCPY = YesPlease
 	NO_MKDTEMP = YesPlease
 	NO_UNSETENV = YesPlease
@@ -834,6 +847,7 @@ ifneq (,$(findstring MINGW,$(uname_S)))
 	NO_C99_FORMAT = YesPlease
 	NO_STRTOUMAX = YesPlease
 	NO_MKDTEMP = YesPlease
+	NO_MKSTEMPS = YesPlease
 	SNPRINTF_RETURNS_BOGUS = YesPlease
 	NO_SVN_TESTS = YesPlease
 	NO_PERL_MAKEMAKER = YesPlease
@@ -853,6 +867,7 @@ ifneq (,$(findstring MINGW,$(uname_S)))
 endif
 ifneq (,$(findstring arm,$(uname_M)))
 	ARM_SHA1 = YesPlease
+	NO_MKSTEMPS = YesPlease
 endif
 
 -include config.mak.autogen
@@ -1011,6 +1026,10 @@ ifdef NO_MKDTEMP
 	COMPAT_CFLAGS += -DNO_MKDTEMP
 	COMPAT_OBJS += compat/mkdtemp.o
 endif
+ifdef NO_MKSTEMPS
+	COMPAT_CFLAGS += -DNO_MKSTEMPS
+	COMPAT_OBJS += compat/mkstemps.o
+endif
 ifdef NO_UNSETENV
 	COMPAT_CFLAGS += -DNO_UNSETENV
 	COMPAT_OBJS += compat/unsetenv.o
diff --git a/compat/mkstemps.c b/compat/mkstemps.c
new file mode 100644
index 0000000..87ebc2a
--- /dev/null
+++ b/compat/mkstemps.c
@@ -0,0 +1,67 @@
+#include "../git-compat-util.h"
+
+#ifndef TMP_MAX
+#define TMP_MAX 16384
+#endif
+
+/* Adapted from libiberty's mkstemp.c. */
+int gitmkstemps(char *pattern, int suffix_len)
+{
+	static const char letters[] =
+		"abcdefghijklmnopqrstuvwxyz"
+		"ABCDEFGHIJKLMNOPQRSTUVWXYZ"
+		"0123456789";
+	static const int num_letters = 62;
+	uint64_t value;
+	struct timeval tv;
+	char *template;
+	size_t len;
+	int fd, count;
+
+	len = strlen(pattern);
+
+	if (len < 6 + suffix_len) {
+		errno = EINVAL;
+		return -1;
+	}
+
+	if (strncmp(&pattern[len - 6 - suffix_len], "XXXXXX", 6)) {
+		errno = EINVAL;
+		return -1;
+	}
+
+	/* Replace pattern's XXXXXX characters with randomness.
+	 * Try TMP_MAX different filenames.
+	 */
+	gettimeofday(&tv, NULL);
+	value = ((size_t)(tv.tv_usec << 16)) ^ tv.tv_sec ^ getpid();
+	template = &pattern[len - 6 - suffix_len];
+	for (count = 0; count < TMP_MAX; ++count) {
+		uint64_t v = value;
+		/* Fill in the random bits. */
+		template[0] = letters[v % num_letters]; v/= num_letters;
+		template[1] = letters[v % num_letters]; v/= num_letters;
+		template[2] = letters[v % num_letters]; v/= num_letters;
+		template[3] = letters[v % num_letters]; v/= num_letters;
+		template[4] = letters[v % num_letters]; v/= num_letters;
+		template[5] = letters[v % num_letters]; v/= num_letters;
+
+		fd = open(pattern, O_CREAT | O_EXCL | O_RDWR, 0600);
+		if (fd > 0)
+			return fd;
+		/* Fatal error (EPERM, ENOSPC etc).
+		 * It doesn't make sense to loop.
+		 */
+		if (errno != EEXIST)
+			break;
+		/* This is a random value.  It is only necessary that
+		 * the next TMP_MAX values generated by adding 7777 to
+		 * VALUE are different with (module 2^32).
+		 */
+		value += 7777;
+	}
+	/* We return the null string if we can't find a unique file name.  */
+	pattern[0] = '\0';
+	errno = EINVAL;
+	return -1;
+}
diff --git a/config.mak.in b/config.mak.in
index 7cce0c1..b6619af 100644
--- a/config.mak.in
+++ b/config.mak.in
@@ -46,6 +46,7 @@ NO_STRTOUMAX=@NO_STRTOUMAX@
 NO_SETENV=@NO_SETENV@
 NO_UNSETENV=@NO_UNSETENV@
 NO_MKDTEMP=@NO_MKDTEMP@
+NO_MKSTEMPS=@NO_MKSTEMPS@
 NO_ICONV=@NO_ICONV@
 OLD_ICONV=@OLD_ICONV@
 NO_DEFLATE_BOUND=@NO_DEFLATE_BOUND@
diff --git a/configure.ac b/configure.ac
index 4e728bc..95dccd4 100644
--- a/configure.ac
+++ b/configure.ac
@@ -676,6 +676,12 @@ GIT_CHECK_FUNC(mkdtemp,
 [NO_MKDTEMP=],
 [NO_MKDTEMP=YesPlease])
 AC_SUBST(NO_MKDTEMP)
+# Define NO_MKSTEMPS if you don't have mkstemps in the C library.
+GIT_CHECK_FUNC(mkstemps,
+[NO_MKSTEMPS=],
+[NO_MKSTEMPS=YesPlease])
+AC_SUBST(NO_MKSTEMPS)
+#
 #
 # Define NO_MMAP if you want to avoid mmap.
 #
diff --git a/git-compat-util.h b/git-compat-util.h
index c7cf2d5..f7217ad 100644
--- a/git-compat-util.h
+++ b/git-compat-util.h
@@ -232,6 +232,11 @@ extern int gitsetenv(const char *, const char *, int);
 extern char *gitmkdtemp(char *);
 #endif
 
+#ifdef NO_MKSTEMPS
+#define mkstemps gitmkstemps
+extern int gitmkstemps(char *, int);
+#endif
+
 #ifdef NO_UNSETENV
 #define unsetenv gitunsetenv
 extern void gitunsetenv(const char *);
-- 
1.6.3.1.169.g33fd

[PATCH v3 2/2] diff: generate prettier filenames when using GIT_EXTERNAL_DIFF

[David Aguilar]

Naturally, prep_temp_blob() did not care about filenames.
As a result, scripts that use GIT_EXTERNAL_DIFF ended up
with filenames such as ".diff_XXXXXX".

This specializes the GIT_EXTERNAL_DIFF code to generate
user-friendly filenames when creating temporary files.

Diffing "name.ext" now generates "XXXXXX_name.ext".

Signed-off-by: David Aguilar <davvid@gmail.com>
Tested-by: Johannes Sixt <j6t@kdbg.org> (Windows)
---
 cache.h                  |    2 ++
 diff.c                   |   45 +++++++++++++++++++++++++++++++++++++--------
 path.c                   |   19 +++++++++++++++++++
 t/t4020-diff-external.sh |    9 +++++++++
 4 files changed, 67 insertions(+), 8 deletions(-)

diff --git a/cache.h b/cache.h
index b8503ad..871c984 100644
--- a/cache.h
+++ b/cache.h
@@ -614,6 +614,8 @@ extern int is_empty_blob_sha1(const unsigned char *sha1);
 
 int git_mkstemp(char *path, size_t n, const char *template);
 
+int git_mkstemps(char *path, size_t n, const char *template, int suffix_len);
+
 /*
  * NOTE NOTE NOTE!!
  *
diff --git a/diff.c b/diff.c
index dcfbcb0..60f07a0 100644
--- a/diff.c
+++ b/diff.c
@@ -1960,12 +1960,37 @@ static void prep_temp_blob(const char *path, struct diff_tempfile *temp,
 			   void *blob,
 			   unsigned long size,
 			   const unsigned char *sha1,
-			   int mode)
+			   int mode,
+			   int pretty_filename)
 {
 	int fd;
 	struct strbuf buf = STRBUF_INIT;
 
-	fd = git_mkstemp(temp->tmp_path, PATH_MAX, ".diff_XXXXXX");
+	if (pretty_filename) {
+		/* Generate "XXXXXX_filename" */
+		struct strbuf pretty_name = STRBUF_INIT;
+		char *basename = ((char*)path) + strlen(path) - 1;
+
+		/* Windows lacks basename() */
+		while(*basename && basename > path) {
+			basename--;
+			if (is_dir_sep(*basename)) {
+				basename++;
+				break;
+			}
+		}
+
+		strbuf_addstr(&pretty_name, "XXXXXX_");
+		strbuf_addstr(&pretty_name, basename);
+
+		fd = git_mkstemps(temp->tmp_path, PATH_MAX,
+			pretty_name.buf, strlen(basename) + 1);
+
+		strbuf_release(&pretty_name);
+	}
+	else {
+		fd = git_mkstemp(temp->tmp_path, PATH_MAX, ".diff_XXXXXX");
+	}
 	if (fd < 0)
 		die("unable to create temp-file: %s", strerror(errno));
 	if (convert_to_working_tree(path,
@@ -1984,7 +2009,8 @@ static void prep_temp_blob(const char *path, struct diff_tempfile *temp,
 }
 
 static struct diff_tempfile *prepare_temp_file(const char *name,
-		struct diff_filespec *one)
+		struct diff_filespec *one,
+		int pretty_filename)
 {
 	struct diff_tempfile *temp = claim_diff_tempfile();
 
@@ -2021,7 +2047,8 @@ static struct diff_tempfile *prepare_temp_file(const char *name,
 				       (one->sha1_valid ?
 					one->sha1 : null_sha1),
 				       (one->sha1_valid ?
-					one->mode : S_IFLNK));
+					one->mode : S_IFLNK),
+				       pretty_filename);
 			strbuf_release(&sb);
 		}
 		else {
@@ -2045,7 +2072,7 @@ static struct diff_tempfile *prepare_temp_file(const char *name,
 		if (diff_populate_filespec(one, 0))
 			die("cannot read data blob for %s", one->path);
 		prep_temp_blob(name, temp, one->data, one->size,
-			       one->sha1, one->mode);
+			       one->sha1, one->mode, pretty_filename);
 	}
 	return temp;
 }
@@ -2071,8 +2098,9 @@ static void run_external_diff(const char *pgm,
 	if (one && two) {
 		struct diff_tempfile *temp_one, *temp_two;
 		const char *othername = (other ? other : name);
-		temp_one = prepare_temp_file(name, one);
-		temp_two = prepare_temp_file(othername, two);
+		int pretty_filename = 1;
+		temp_one = prepare_temp_file(name, one, pretty_filename);
+		temp_two = prepare_temp_file(othername, two, pretty_filename);
 		*arg++ = pgm;
 		*arg++ = name;
 		*arg++ = temp_one->name;
@@ -3574,8 +3602,9 @@ static char *run_textconv(const char *pgm, struct diff_filespec *spec,
 	const char **arg = argv;
 	struct child_process child;
 	struct strbuf buf = STRBUF_INIT;
+	int pretty_filename = 0;
 
-	temp = prepare_temp_file(spec->path, spec);
+	temp = prepare_temp_file(spec->path, spec, pretty_filename);
 	*arg++ = pgm;
 	*arg++ = temp->name;
 	*arg = NULL;
diff --git a/path.c b/path.c
index 8a0a674..090b490 100644
--- a/path.c
+++ b/path.c
@@ -140,6 +140,25 @@ int git_mkstemp(char *path, size_t len, const char *template)
 }
 
 
+
+/* git_mkstemps() - create tmp file with suffix honoring TMPDIR variable. */
+int git_mkstemps(char *path, size_t len, const char *template, int suffix_len)
+{
+	const char *tmp;
+	size_t n;
+
+	tmp = getenv("TMPDIR");
+	if (!tmp)
+		tmp = "/tmp";
+	n = snprintf(path, len, "%s/%s", tmp, template);
+	if (len <= n) {
+		errno = ENAMETOOLONG;
+		return -1;
+	}
+	return mkstemps(path, suffix_len);
+}
+
+
 int validate_headref(const char *path)
 {
 	struct stat st;
diff --git a/t/t4020-diff-external.sh b/t/t4020-diff-external.sh
index 0720001..4ea42e0 100755
--- a/t/t4020-diff-external.sh
+++ b/t/t4020-diff-external.sh
@@ -136,6 +136,15 @@ test_expect_success 'GIT_EXTERNAL_DIFF with more than one changed files' '
 	GIT_EXTERNAL_DIFF=echo git diff
 '
 
+test_expect_success 'GIT_EXTERNAL_DIFF generates pretty paths' '
+	touch file.ext &&
+	git add file.ext &&
+	echo with extension > file.ext &&
+	GIT_EXTERNAL_DIFF=echo git diff file.ext | grep ......_file\.ext &&
+	git update-index --force-remove file.ext &&
+	rm file.ext
+'
+
 echo "#!$SHELL_PATH" >fake-diff.sh
 cat >> fake-diff.sh <<\EOF
 cat $2 >> crlfed.txt
-- 
1.6.3.1.169.g33fd

Re: [PATCH v3 1/2] compat: add a mkstemps() compatibility function

[Antriksh Pany]

Hi David

I think I see a few minor issues with the code dealing with random six
letters generation. Please correct me if I am wrong.

On Thu, May 28, 2009 at 3:13 PM, David Aguilar <davvid@gmail.com> wrote:
> mkstemps() is a BSD extension so provide an implementation
> for cross-platform use.
>
> Signed-off-by: David Aguilar <davvid@gmail.com>
> Tested-by: Johannes Sixt <j6t@kdbg.org> (Windows)
> ---
>  Makefile          |   19 +++++++++++++++
>  compat/mkstemps.c |   67 +++++++++++++++++++++++++++++++++++++++++++++++++++++
>  config.mak.in     |    1 +
>  configure.ac      |    6 ++++
>  git-compat-util.h |    5 ++++
>  5 files changed, 98 insertions(+), 0 deletions(-)
>  create mode 100644 compat/mkstemps.c
>
> diff --git a/Makefile b/Makefile
> index eaae45d..a70b5f0 100644
> --- a/Makefile
> +++ b/Makefile
> @@ -52,6 +52,8 @@ all::
>  #
>  # Define NO_MKDTEMP if you don't have mkdtemp in the C library.
>  #
> +# Define NO_MKSTEMPS if you don't have mkstemps in the C library.
> +#
>  # Define NO_SYS_SELECT_H if you don't have sys/select.h.
>  #
>  # Define NO_SYMLINK_HEAD if you never want .git/HEAD to be a symbolic link.
> @@ -636,10 +638,12 @@ EXTLIBS =
>
>  ifeq ($(uname_S),Linux)
>        NO_STRLCPY = YesPlease
> +       NO_MKSTEMPS = YesPlease
>        THREADED_DELTA_SEARCH = YesPlease
>  endif
>  ifeq ($(uname_S),GNU/kFreeBSD)
>        NO_STRLCPY = YesPlease
> +       NO_MKSTEMPS = YesPlease
>        THREADED_DELTA_SEARCH = YesPlease
>  endif
>  ifeq ($(uname_S),UnixWare)
> @@ -651,6 +655,7 @@ ifeq ($(uname_S),UnixWare)
>        SHELL_PATH = /usr/local/bin/bash
>        NO_IPV6 = YesPlease
>        NO_HSTRERROR = YesPlease
> +       NO_MKSTEMPS = YesPlease
>        BASIC_CFLAGS += -Kthread
>        BASIC_CFLAGS += -I/usr/local/include
>        BASIC_LDFLAGS += -L/usr/local/lib
> @@ -674,6 +679,7 @@ ifeq ($(uname_S),SCO_SV)
>        SHELL_PATH = /usr/bin/bash
>        NO_IPV6 = YesPlease
>        NO_HSTRERROR = YesPlease
> +       NO_MKSTEMPS = YesPlease
>        BASIC_CFLAGS += -I/usr/local/include
>        BASIC_LDFLAGS += -L/usr/local/lib
>        NO_STRCASESTR = YesPlease
> @@ -702,6 +708,7 @@ ifeq ($(uname_S),SunOS)
>        NO_MEMMEM = YesPlease
>        NO_HSTRERROR = YesPlease
>        NO_MKDTEMP = YesPlease
> +       NO_MKSTEMPS = YesPlease
>        OLD_ICONV = UnfortunatelyYes
>        ifeq ($(uname_R),5.8)
>                NO_UNSETENV = YesPlease
> @@ -724,6 +731,7 @@ ifeq ($(uname_O),Cygwin)
>        NO_D_INO_IN_DIRENT = YesPlease
>        NO_STRCASESTR = YesPlease
>        NO_MEMMEM = YesPlease
> +       NO_MKSTEMPS = YesPlease
>        NO_SYMLINK_HEAD = YesPlease
>        NEEDS_LIBICONV = YesPlease
>        NO_FAST_WORKING_DIRECTORY = UnfortunatelyYes
> @@ -767,11 +775,13 @@ ifeq ($(uname_S),NetBSD)
>        BASIC_LDFLAGS += -L/usr/pkg/lib $(CC_LD_DYNPATH)/usr/pkg/lib
>        THREADED_DELTA_SEARCH = YesPlease
>        USE_ST_TIMESPEC = YesPlease
> +       NO_MKSTEMPS = YesPlease
>  endif
>  ifeq ($(uname_S),AIX)
>        NO_STRCASESTR=YesPlease
>        NO_MEMMEM = YesPlease
>        NO_MKDTEMP = YesPlease
> +       NO_MKSTEMPS = YesPlease
>        NO_STRLCPY = YesPlease
>        NO_NSEC = YesPlease
>        FREAD_READS_DIRECTORIES = UnfortunatelyYes
> @@ -787,12 +797,14 @@ endif
>  ifeq ($(uname_S),GNU)
>        # GNU/Hurd
>        NO_STRLCPY=YesPlease
> +       NO_MKSTEMPS = YesPlease
>  endif
>  ifeq ($(uname_S),IRIX64)
>        NO_IPV6=YesPlease
>        NO_SETENV=YesPlease
>        NO_STRCASESTR=YesPlease
>        NO_MEMMEM = YesPlease
> +       NO_MKSTEMPS = YesPlease
>        NO_STRLCPY = YesPlease
>        NO_SOCKADDR_STORAGE=YesPlease
>        SHELL_PATH=/usr/gnu/bin/bash
> @@ -805,6 +817,7 @@ ifeq ($(uname_S),HP-UX)
>        NO_SETENV=YesPlease
>        NO_STRCASESTR=YesPlease
>        NO_MEMMEM = YesPlease
> +       NO_MKSTEMPS = YesPlease
>        NO_STRLCPY = YesPlease
>        NO_MKDTEMP = YesPlease
>        NO_UNSETENV = YesPlease
> @@ -834,6 +847,7 @@ ifneq (,$(findstring MINGW,$(uname_S)))
>        NO_C99_FORMAT = YesPlease
>        NO_STRTOUMAX = YesPlease
>        NO_MKDTEMP = YesPlease
> +       NO_MKSTEMPS = YesPlease
>        SNPRINTF_RETURNS_BOGUS = YesPlease
>        NO_SVN_TESTS = YesPlease
>        NO_PERL_MAKEMAKER = YesPlease
> @@ -853,6 +867,7 @@ ifneq (,$(findstring MINGW,$(uname_S)))
>  endif
>  ifneq (,$(findstring arm,$(uname_M)))
>        ARM_SHA1 = YesPlease
> +       NO_MKSTEMPS = YesPlease
>  endif
>
>  -include config.mak.autogen
> @@ -1011,6 +1026,10 @@ ifdef NO_MKDTEMP
>        COMPAT_CFLAGS += -DNO_MKDTEMP
>        COMPAT_OBJS += compat/mkdtemp.o
>  endif
> +ifdef NO_MKSTEMPS
> +       COMPAT_CFLAGS += -DNO_MKSTEMPS
> +       COMPAT_OBJS += compat/mkstemps.o
> +endif
>  ifdef NO_UNSETENV
>        COMPAT_CFLAGS += -DNO_UNSETENV
>        COMPAT_OBJS += compat/unsetenv.o
> diff --git a/compat/mkstemps.c b/compat/mkstemps.c
> new file mode 100644
> index 0000000..87ebc2a
> --- /dev/null
> +++ b/compat/mkstemps.c
> @@ -0,0 +1,67 @@
> +#include "../git-compat-util.h"
> +
> +#ifndef TMP_MAX
> +#define TMP_MAX 16384
> +#endif
> +
> +/* Adapted from libiberty's mkstemp.c. */
> +int gitmkstemps(char *pattern, int suffix_len)
> +{
> +       static const char letters[] =
> +               "abcdefghijklmnopqrstuvwxyz"
> +               "ABCDEFGHIJKLMNOPQRSTUVWXYZ"
> +               "0123456789";
> +       static const int num_letters = 62;
> +       uint64_t value;
> +       struct timeval tv;
> +       char *template;
> +       size_t len;
> +       int fd, count;
> +
> +       len = strlen(pattern);
> +
> +       if (len < 6 + suffix_len) {
> +               errno = EINVAL;
> +               return -1;
> +       }
> +
> +       if (strncmp(&pattern[len - 6 - suffix_len], "XXXXXX", 6)) {
> +               errno = EINVAL;
> +               return -1;
> +       }
> +
> +       /* Replace pattern's XXXXXX characters with randomness.
> +        * Try TMP_MAX different filenames.
> +        */
> +       gettimeofday(&tv, NULL);
> +       value = ((size_t)(tv.tv_usec << 16)) ^ tv.tv_sec ^ getpid();
> +       template = &pattern[len - 6 - suffix_len];
> +       for (count = 0; count < TMP_MAX; ++count) {
> +               uint64_t v = value;
> +               /* Fill in the random bits. */
> +               template[0] = letters[v % num_letters]; v/= num_letters;
> +               template[1] = letters[v % num_letters]; v/= num_letters;
> +               template[2] = letters[v % num_letters]; v/= num_letters;
> +               template[3] = letters[v % num_letters]; v/= num_letters;
> +               template[4] = letters[v % num_letters]; v/= num_letters;
> +               template[5] = letters[v % num_letters]; v/= num_letters;

v is divided 6 times by 62, and a decent amount of randomness shall be
ensured only if 'value' is usually greater than 62^6.

If it is assumed to be ok for 'v' to become zero in the last one odd
steps sometimes (in fact, _frequently_), then you may ignore what I am
pointing out here.

1. On a couple of systems that I checked on, all variables/types on
the right hand side of
      value = ((size_t)(tv.tv_usec << 16)) ^ tv.tv_sec ^ getpid();
are 4 bytes long. So 'value' is ultimately going to be assigned a
number that fits in 4 bytes, ie value < 2^32 (at least on a few
systems).
The systems that I checked on have these `uname -a` outputs:
-   Linux host 2.6.9-55.ELsmp #1 SMP Fri Apr 20 17:03:35 EDT 2007 i686
i686 i386 GNU/Linux
-   Linux host 2.4.21-50.ELsmp #1 SMP Tue May 8 17:18:29 EDT 2007 i686
i686 i386 GNU/Linux

Also, even if size_t were 64 bits, typecasting _after_ the shift does
not help much. And given the uncertainty about sizeof (size_t), we
could use:
  ((uint64_t) tv.tv_usec) << 16)


2. tv_usec has a decimal value range of 0-999999  (10^6 usec make 1
sec). Which means that tv_usec fits completely in 20 bits (or less).
(tv_usec << 16) yields a number that fits completely in 36 bits (or
less). Max value of this number is 999999 * 2^16, or for convenience,
about M = 10^6 * 2^16. This number (in the range of 0 to M) goes on to
be divided by D=62^6. Also, M > D. Thus, there is about a  D / M * 100
~ 87 % probability of the division M / D working out to be zero. i.e,
in 87% cases, the last division (v/= num_letters) will cause 'v' to
become 0.
Therefore, I think an additional shift of 7 or more bits will help in
improving randomness of template[5]. I would suggest left shifting by
24 bits. This seemed ok in some tests I did.


Combining #1 and #2, I guess we could have the computation of 'value' as:
      value = (((uint64_t)tv.tv_usec) << 24) ^ tv.tv_sec ^ getpid();


- Antriksh

Re: [PATCH v3 1/2] compat: add a mkstemps() compatibility function

[Johannes Sixt]

Antriksh Pany schrieb:
> 2. tv_usec has a decimal value range of 0-999999  (10^6 usec make 1
> sec). Which means that tv_usec fits completely in 20 bits (or less).
> (tv_usec << 16) yields a number that fits completely in 36 bits (or
> less). Max value of this number is 999999 * 2^16, or for convenience,
> about M = 10^6 * 2^16. This number (in the range of 0 to M) goes on to
> be divided by D=62^6. Also, M > D. Thus, there is about a  D / M * 100
> ~ 87 % probability of the division M / D working out to be zero. i.e,
> in 87% cases, the last division (v/= num_letters) will cause 'v' to
> become 0.

But this value (after the 6th division, mind you) is not used. v is used
the last time after the 5th division; at this time it still carries enough
randomness: 62^5 < 2^32.

-- Hannes

Re: [PATCH v3 1/2] compat: add a mkstemps() compatibility function

[Antriksh Pany]

Oh ok, didn't realize the sixth division was not used. In which case,
we needn't have the sixth division.

Also, 62^5 compared to 2^32 still gives a ~ 21% probability of the
fifth division becoming zero, fairly high perhaps.

The suggestion to change the code to
   ((uint64_t) tv.tv_usec) << 16)
might help here. This would still leave about a 1.4% chance of v
becoming zero, higher than (being pedantic!) what perfect randomness
would require. Livable. But (just curious), do we see any loss in
shifting a few more positions?

- Antriksh


On Tue, Jun 2, 2009 at 6:35 PM, Johannes Sixt <j.sixt@viscovery.net> wrote:
> Antriksh Pany schrieb:
>> 2. tv_usec has a decimal value range of 0-999999  (10^6 usec make 1
>> sec). Which means that tv_usec fits completely in 20 bits (or less).
>> (tv_usec << 16) yields a number that fits completely in 36 bits (or
>> less). Max value of this number is 999999 * 2^16, or for convenience,
>> about M = 10^6 * 2^16. This number (in the range of 0 to M) goes on to
>> be divided by D=62^6. Also, M > D. Thus, there is about a  D / M * 100
>> ~ 87 % probability of the division M / D working out to be zero. i.e,
>> in 87% cases, the last division (v/= num_letters) will cause 'v' to
>> become 0.
>
> But this value (after the 6th division, mind you) is not used. v is used
> the last time after the 5th division; at this time it still carries enough
> randomness: 62^5 < 2^32.
>
> -- Hannes
>
>

Re: [PATCH v3 1/2] compat: add a mkstemps() compatibility function

[David Aguilar]

On Wed, Jun 03, 2009 at 09:29:06PM +0530, Antriksh Pany wrote:
> Oh ok, didn't realize the sixth division was not used. In which case,
> we needn't have the sixth division.
> 
> Also, 62^5 compared to 2^32 still gives a ~ 21% probability of the
> fifth division becoming zero, fairly high perhaps.
> 
> The suggestion to change the code to
>    ((uint64_t) tv.tv_usec) << 16)
> might help here. This would still leave about a 1.4% chance of v
> becoming zero, higher than (being pedantic!) what perfect randomness
> would require. Livable. But (just curious), do we see any loss in
> shifting a few more positions?
> 
> - Antriksh


Just a small note -- our mkstemps was adapted from
libiberty's mkstemps.c (gcc-4.something), so if there is
indeed something to fix here then we should let them know.

I looked in libiberty since I assumed that the code in there
was battle-tested and stable.

I feel okay with it as-is, but that's just me.



> On Tue, Jun 2, 2009 at 6:35 PM, Johannes Sixt <j.sixt@viscovery.net> wrote:
> > Antriksh Pany schrieb:
> >> 2. tv_usec has a decimal value range of 0-999999  (10^6 usec make 1
> >> sec). Which means that tv_usec fits completely in 20 bits (or less).
> >> (tv_usec << 16) yields a number that fits completely in 36 bits (or
> >> less). Max value of this number is 999999 * 2^16, or for convenience,
> >> about M = 10^6 * 2^16. This number (in the range of 0 to M) goes on to
> >> be divided by D=62^6. Also, M > D. Thus, there is about a  D / M * 100
> >> ~ 87 % probability of the division M / D working out to be zero. i.e,
> >> in 87% cases, the last division (v/= num_letters) will cause 'v' to
> >> become 0.
> >
> > But this value (after the 6th division, mind you) is not used. v is used
> > the last time after the 5th division; at this time it still carries enough
> > randomness: 62^5 < 2^32.
> >
> > -- Hannes
> >
> >

-- 
		David